A time-resolved imaging-based CRISPRi screening method  Daniel Camsund1*, Michael J. Lawson1,2*#, Jimmy Larsson1, Daniel Jones1, Spartak Zikrin1, David                       Fange1 & Johan Elf1#  1 Department of Cell and Molecular Biology, SciLifeLab, Uppsala University, 75124 Uppsala. 2 Current Address: Division of Biology and Biological Engineering, California Institute of                       Technology, Pasadena, CA, USA.  * Equal contribution # Corresponding author: johan.elf@icm.uu.se, lawsonjmichael@gmail.com     

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted August 27, 2019. . https://doi.org/10.1101/747758doi: bioRxiv preprint

Abstract  Our ability to connect genotypic variation to biologically important phenotypes has been seriously                         limited by the gap between live cell microscopy and library-scale genomic engineering. Specifically,                         this has restricted studies of intracellular dynamics to one strain at a time and thus, generally, to the                                   impact of genes with known function. Here we show how in situ genotyping of a library of E. coli                                     strains after time-lapse imaging in a microfluidic device overcomes this problem. We determine how                           235 different CRISPR interference (CRISPRi) knockdowns impact the coordination of the replication                       and division cycles of E. coli by monitoring the location of replication forks throughout on average                               >500 cell cycles per knockdown. The single-cell time-resolved assay allows us to determine the                           distribution of single-cell growth rates, cell division sizes, and replication initiation volumes.                       Subsequent in situ genotyping allows us to map each phenotype distribution to a specific genetic                             perturbation in order to determine which genes are important for cell cycle control. The technology                             presented in this study enables genome-scale screens of virtually all live-cell microscopy assays and,                           therefore, constitutes a qualitatively new approach to cellular biophysics.   

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted August 27, 2019. . https://doi.org/10.1101/747758doi: bioRxiv preprint

Introduction  The last decade has shown remarkable development in genome engineering, fronted by applications of                           Cas9-mediated gene targeting 1,2. In combination with inexpensive large-scale DNA oligonucleotide                     synthesis, these techniques make it possible to generate pool-synthesized cell strain libraries with                         specific perturbations genome-wide 3,4. More recently, methods have been developed for screening                       Cas9 genome edited libraries by sorting the library members based on the expression of a fluorescent                               protein and then sequencing cells with similar expression level 5 or using single-cell RNAseq on                             libraries of CRISPR interference (CRISPRi) perturbations to determine the state of the transcriptome                         for individual cells 6–8. These methods are however blind to cellular dynamics and intracellular                           localization of relevant molecules.  The progress in genome-scale engineering and expression perturbation has been accompanied by                       equally impressive developments in microscopy and microfabrication, which enable characterization                   of complex phenotypes at high temporal and spatial resolution in living cells under well-controlled                           conditions 9–15. While the power of these methods enables deep insight into cellular biophysics, the                             limitation of working with one strain at a time prohibits studying the impact of genes whose function                                 is not already, at least to some degree, known. Given the rapid progress within the previously separate                                 fields of imaging and genomics, the lack of efficient techniques for time-resolved single-cell                         phenotyping of pool-synthesized genetic strain libraries constitutes a severe bottleneck in biological                       research. We have recently proposed a tentative solution to the problem 16 that we now demonstrate                             scaled to hundreds of genes for tens of thousands of single cells, converting the concept into a                                 practical tool. We use the method to perform a large-scale screen of complex phenotypes to identify    the regulatory elements of replication-division coordination in bacterial cells.

Results Overview of the method The heart of our method is a microfluidic device that enables both high-resolution dynamic                           phenotyping and subsequent in situ genotyping of the individual strains. The microfluidics approach                         allows us to keep the bacteria in a constant state of exponential growth over hundreds of generations                                 while imaging them at high resolution. The fluidic device 13 is an adaption of the “mother machine                                 chip” 15 where each cell trap has a 300 nm constriction at the end that enables fast loading, media and                                       probe exchange (see the left side of Fig. 1 for a schematic of the chip). Each strain occupies a defined                                       space in the fluidic chip, but the genotypes of the different strains are unknown at the time of                                   phenotyping. After the phenotypes have been determined, the cells are chemically fixed in the chip,                             and the genotypes are optically inferred by sequential fluorescence in situ hybridization (FISH) to a                             barcode (Fig. 1). We refer to the technique as DuMPLING (Dynamic u-fluidic Microscopy                         Phenotyping of a Library before IN situ Genotyping).   

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted August 27, 2019. . https://doi.org/10.1101/747758doi: bioRxiv preprint

  

Figure 1 Assay workflow. 1. The bacteria in the CRISPRi library contain pool-synthesized                         plasmids each expressing a barcode and a corresponding sgRNA (top) for repressing a                         specific gene in the E. coli chromosome. The library of cells is loaded into a fluidic device                                 where each strain occupies a spatially separated trap (left) and where 2. the cells can be                               monitored with highly sensitive time-lapse fluorescence microscopy for hours or days. 3.                       After phenotyping the cells are fixed, and the identities of the strains are revealed by                             sequential FISH probing for the barcodes.   

 We used the DuMPLING technique to characterize the coordination of the replication and division                           cycle in E. coli by tracking the chromosome replication forks throughout the cell cycle in a CRISPRi                                 library. Replication initiation in E. coli is triggered at a fixed volume per chromosome 17,18                             independent of growth rate, but the underlying molecular mechanism is largely unknown. In this                           work, replication initiation was studied directly by observing a strain with a chromosomally integrated                           seqA-yfp fusion. SeqA binds hemimethylated DNA in the wake of the replication machinery and can                             thus be used to track replication foci (Fig. 2). In addition, the cell size at division and replication                                   initiation was determined using phase contrast imaging. By imaging hundreds of cell cycles for each                             CRISPRi perturbation, we aimed to identify which genes are involved in setting the accuracy of the                               initiation volume. This screen could not have been performed without monitoring the dynamics of                           replication initiation directly in individual cells.    Implementation and Analysis  We constructed the library in a host strain of MG1655 with a seqA-yfp fusion for tracking the                                 replication forks, dCas9 expressed from an anhydrotetracycline (aTc) inducible promoter for                     CRISPRi, and a T7 RNA polymerase (T7 RNApol) expressed from an arabinose-inducible promoter                 for barcode RNA expression. Each of the 235 library members expressed a unique single guide RNA                          (sgRNA), directing dCas9 to bind and repress a specific gene (Online Methods 1.1.2). The library                             included all known, non-lethal cell cycle-related targets as well as 38 y-genes 19, 28 of which are                                 largely uncharacterized or have an unknown function. By inducing the dCas9 expression with 1 pg/ul                             aTc, the target genes were downregulated between 10-100 times (Supplementary Text 2.2). Each                         plasmid also encoded a 20 bp-barcode sequence that could be expressed as RNA from an inducible T7                                 promoter. The barcode was uniquely coordinated with the sgRNA (Online Methods 1.1.4) sequence to                           identify which sgRNA was expressed in which strain.  

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted August 27, 2019. . https://doi.org/10.1101/747758doi: bioRxiv preprint

 We loaded 40 pool synthesized strains at a time into the microfluidic device in order to phenotype on                                   average 562 generations per strain in an 8 h experiment. Before phenotyping, the bacteria were grown                               with dCas9 induced for 9 hours in order to establish steady state phenotypes. This time also allowed                                 the mother cell at the end of the cell trap to divide enough times to set the genotype of the whole trap.                                           To achieve sufficient time resolution, we imaged each position (30 traps per position) for 20 ms every                                 minute in the phase contrast channel and 300 ms every two minutes in the fluorescence channel (514                                 nm@5.3 W/cm2). This limited us to 90 positions, or 2700 cell traps per experiment. This also gives                                enough traps per strain to account for uneven representation of individual genotypes in the library                             (Fig. S1). The limitation for how many strains we can analyze simultaneously is currently set by the                                 number of traps we can image during phenotyping (e.g. how fast we can image and move the stage)                                   and not by how many strains can be made or genotyped in parallel.   After phenotyping, the barcode RNAs were expressed by T7 RNApol induction and the cells were                             subsequently fixed and permeabilized. The strains were identified in situ by sequential FISH probing                           in four colors; each round of probing identifies the positions of four unique strains. Each round was                                 completed in less than 30 mins, since no stripping of probes was required, and all strains could be                                   identified within 6 hours following fixation. In complementary experiments where we loaded all                         strains simultaneously, we used combinatorial FISH probing which can identify NR genotypes in R                           rounds of N colors, but this approach is more time-consuming at this library-scale since the probe                               stripping is slow and the readout probe concentrations are lower (see Supplementary Text 2.3 for                             additional information about advantages and drawbacks of different genotyping methods in the PDMS                         chip).    

   

Figure 2 Analysis. A. Cartoon of cells growing in a microfluidic chip. B. Example                           kymographs in phase contrast (top) and fluorescence (bottom), with automated cell                     

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted August 27, 2019. . https://doi.org/10.1101/747758doi: bioRxiv preprint

segmentation (yellow) or detection of SeqA-YFP clusters (red circles), respectively,                   overlayed. C. The time series of cell segmentation and SeqA-YFP clusters are combined to                           generate single cell trajectories across cell divisions and chromosome replications. 

 After connecting the genotypes to the phenotypes through their spatial position, we analyzed how the                             repression of individual genes affected the cell size and growth rate (Fig 2B top) as well as                                 coordination of replication with the division cycle (Fig 2B bottom). The kymograph for one of the                               21700 analysed cell traps is displayed in Fig 2C. The bacteria were segmented using the POE method                                 20 and tracked by the Baxter algorithm 21. The SeqA-YFP foci were detected by a wavelet based                                 method 22 and the replication forks were tracked simultaneously through the generations using the                           u-track algorithm 23 (Fig 2C). Each experiment, including 2700 cell traps that were imaged every                             minute for 8 hours, resulted in 220 Gb of image data and took 2 hours to analyze on 45 cores using                                         customized parallelized image analysis routines.          

 Figure 3 Phenotypic data averaged for each genotype. A-C Two dimensional plots in                         phenotype space, CRISPRi knockdowns with significant deviation from the reference control                     strain (ref) are labeled by the name of the targeted gene. Outlier dots have been classified and                                 labelled as follows: large birth size (green), small birth size (cyan), large initiation size                           (orange), small initiation size (yellow), no foci (grey) and undefinable initiation size (purple).                         As these properties are not all mutually exclusive, dots may by multicolored. A. Horizontal:                           Average normalized birth size. Vertical: Average normalized initiation size. B. Horizontal:                     Average normalized birth size. Vertical: Average normalized growth rate. C. Horizontal:                     Average normalized growth rate. Vertical: Average normalized initiation size. D. Control                     experiment showing that the relative growth rate derived from the DuMPLING experiments                       correlate well with that from a pooled competition assay (correlation = 0.7). E. Fork                           distribution plot. Horizontal is SeqA-YFP cluster location along the long axis of the cell(from                           old cell pole to new), vertical is cell size, color indicates the probability of finding a                               

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted August 27, 2019. . https://doi.org/10.1101/747758doi: bioRxiv preprint

replication fork at a given position along the cell axis and a given cell size. Initiation size                                 corresponds to the average of individually tracked replication forks. F. Average normalized                       initiation size fit to bulk replication forks (vertical) compared with the average of single cell                             initiation size estimates (correlation = 0.76). 

  The impact of perturbations on the E. coli cell cycle In Fig 3A-C we show comparisons of the average growth rates and cell sizes at division and                                 replication initiation respectively for the 215 strains for which we obtain data from a minimum of 5                                 independent cell traps and 40 complete cell cycles. The genotypes that are substantially different from                             the reference control strain in replicate experiments (ref, see Online Methods 1.1.4 for strain details)                             are indicated by the name of the sgRNA targeted gene. The sizes at division and initiation can get                                   both bigger and smaller than ref, whereas growth rates typically only get smaller. The deviations from                               ref are mostly uncorrelated between the properties, with two notable exceptions: the tol-pal cluster is                             smaller and the fis, diaA-cluster is larger in both initiation and birth volumes (Fig. 3A).    As a control for correct genotyping and cell segmentation, the average growth rates obtained from the                               single-cell time-lapse imaging compare well to the corresponding bulk experiment (Fig. 3D). To make                           bulk estimates of the growth rates, we performed a competition assay of the whole library in liquid                                 culture and determined the time-dependent relative abundance of each genotype by next generation                         sequencing (NGS) (Fig 3D, assay details in Supplementary Text 2.1). Additional controls are                         described in the Supplementary Text 2.4.1-2.4.6. For example, we compared the phenotypes of                         selected strains from the DuMPLING screen to those obtained with the corresponding knock-outs or                           specific sgRNA knock-downs constructed and measured one strain at a time (Supplemental Text                         2.4.4). The good agreement in phenotypes implies that the genotyping is robust, although there were                             notable exceptions with off-target effects or where synthesis errors in the sgRNA coding sequence                           were selected for.    An information-rich way to simultaneously characterize replication and cell cycle processes in                       individual strains is the fork distribution plot, which shows the probability of finding a replication fork                               in a specific position in the cell (horizontal axis) for cells of a given size (vertical axis). The                                   distribution for the ref strain is shown in Fig 3E. The data in the fork distribution plots can be used to                                         estimate average initiation volume (see Fig. 3E, Online Methods 1.3.1 and 17). The good agreement                             between this bulk estimate and the average of estimates obtained from individual cells (see Fig. 3F for                                 comparison and Online Methods 1.3.1 for estimation of a single cell initiation size) acts as a control                     for our ability to accurately estimate initiation size for individual cell trajectories.   

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted August 27, 2019. . https://doi.org/10.1101/747758doi: bioRxiv preprint

 Figure 4 Fork distribution plots and time-resolved phenotypes. A. Plots for all                       knockdowns that had a minimum of 5 independent cell traps and 40 whole cell cycles. Color                               labeling indicates the classification of phenotypes. Fork distribution plots are laid out as in                           Fig. 3E. White dashed lines: Birth/division size. Red dashed lines: replication initiation size,                         average of individually tracked replication forks. B-G: Examples of each classification from                       A (NB: it is possible for a gene to belong to more than one classification). B. Large division                                   size (only birth size in figure). C. Small division size (only division size in figure). D. Large                                 initiation size. E. Small initiation size. F. No SeqA-YFP foci. G. Replication initiation size                           not definable. Fork distribution plots for a replica experiment is shown in figures S2-S7. 

 Fig. 4A show the fork distribution plots for all genotypes from one experiment that met the criteria                                 described in the figure legend. Most of these are similar to the unperturbed ref (Fig 3E), but a number                                     of variant classes can be identified: Fig 4B (fis, dedD etc.) represents strains with large division size                                 and Fig 4C (clpP, nfh, pal etc.) strains with small division size. Correspondingly Fig 4D (fis, ihf etc.)                                   represents strains with large initiation size and Fig 4E (tol, pal etc.) represents strains with small                               initiation size. The strains in Fig 4F (seqA, dam and damX) have no SeqA-YFP foci, which makes                                 sense since dam encodes the DNA methylase and SeqA only binds hemimethylated DNA. damX is                             located upstream of dam in the same operon and its repression will also downregulate dam. Fig 4G                                 (hda, ihf etc.) represents strains for which the replication initiation size is undefined, suggesting that                             there may be significant cell-to-cell variability in the initiation size and consequently, that these genes                             are important for regulation of DNA replication.     

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted August 27, 2019. . https://doi.org/10.1101/747758doi: bioRxiv preprint

 Figure 5 Cell-to-cell phenotypic variation for each genotype. Top row: Each plot has the                           coefficient of variation (CV) on the vertical axis and mean on the horizontal axis, both                             normalized by ref (red dot at [1,1] in each plot), for A. growth rate, B. birth size and C.                                     initiation size. Bottom row: In D-F we show probability density (vertical axis) as a function                             of D. growth rate, E. birth size and F. initiation size for selected genotypes.   

By pooling data from different cells of the same genotype, we lose information about cell-to-cell                             variation for the different genotypes. As a first-order description of the cell-to-cell variation, we                           plotted the coefficient of variation (CV) against the corresponding average in growth rate, birth size                             and initiation size for each genotype in Fig 5A-C. Corresponding illustrative examples of the full                             distributions are also given for selected genotypes in Fig 5D-F. For birth size most perturbations give                               larger average as well as CV (Fig. 5B). Interestingly fis and dedD do however increase the average                                 size without an increase in CV. In a few cases the CV increases much more than the change in the                                       average. For example, hda repression gives a 80% increase in variation of the replication initiation                             size with only a 10% reduction in the average (Fig. 5C), suggesting an important regulatory role. With                                 respect to cell-to-cell variation in growth rate, we observe a striking trend that gene knockdowns that                               reduce growth rate also give rise to large cell-to-cell variation (Fig. 5A). A plausible explanation is                               that the repression makes cell growth limited in only one factor and that stochastic fluctuations in this                                 factor alone directly impact the growth rate.  

Discussion  Circling back to the original question of which genes are important for triggering replication at a fixed                                 cell size, a number of candidates stand out in Fig 4G and 5C. If we exclude genes with well                                     documented and clearly unrelated functions, we end up with hda, diaA, ihfA, ihfB, fis, yjgH and yeeS.                                  diaA is previously known for its direct interaction with DnaA at replication initiation 24. hda, ihfA,                               ihfB, and fis are all associated with DnaA-ATP to ADP cycling 25. Further characterization of yjgH                               shows that the observed phenotype is due to an off-target effect of the sgRNA (Supplementary Text                               2.4.3). The remaining uncharacterized gene that shows an evident perturbation in the accuracy of                           replication is yeeS. This gene deserves more specific study, although there is a chance that the                               perturbation is mediated by the CbtA toxin 26 that is encoded on the same transcript as yeeS. We also                                     note that downregulation of pgsA does not show a perturbation in replication initiation although the                             

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted August 27, 2019. . https://doi.org/10.1101/747758doi: bioRxiv preprint

gene product is key in the synthesis of fatty acids, which some studies have implicated in DnaA-ADP                                 to ATP conversion at the membrane 27. Thus, overall, our assay supports models for replication                      initiation control based on DnaA-ATP to ADP cycling through the regulatory inactivation of DnaA (RIDA) mechanism mediated by Hda 28 as well as at the DnaA-reactivating sequences (DARS) 29 binding Fis and Ihf and datA locus dependent DnaA-ATP hydrolysis (DDAH) 30 also regulated by Ihf. The DuMPLING approach, here deployed to identify the key regulators of the E. coli cell cycle, can                                 be used to study all sorts of complex dynamic phenotypic traits that require sensitive or time-lapse                               microscopy for characterization. Plausible extensions include studies of gene expression dynamics in                       response to recoded promoter sequences or other genetic regulatory elements; live cell enzymatic                         assays as a function of active site residue mutations; as well as interaction partner screens that alter                                 the intracellular location of a labeled reporter molecule in response to all possible gene knockdowns.     

References 1. Jinek, M. et al. A Programmable Dual-RNA–Guided DNA Endonuclease in Adaptive Bacterial 

Immunity. Science 337, 816–821 (2012). 

2. Adli, M. The CRISPR tool kit for genome editing and beyond. Nat. Commun. 9, 1911 (2018). 

3. Garst, A. D. et al. Genome-wide mapping of mutations at single-nucleotide resolution for 

protein, metabolic and genome engineering. Nat. Biotechnol. 35, 48–55 (2016). 

4. Wang, H. H. et al. Programming cells by multiplex genome engineering and accelerated 

evolution. Nature 460, 894–898 (2009). 

5. Rajagopal, N. et al. High-throughput mapping of regulatory DNA. Nat. Biotechnol. 34, 167–174 

(2016). 

6. Dixit, A. et al. Perturb-Seq: Dissecting Molecular Circuits with Scalable Single-Cell RNA 

Profiling of Pooled Genetic Screens. Cell 167, 1853–1866.e17 (2016). 

7. Adamson, B. et al. A Multiplexed Single-Cell CRISPR Screening Platform Enables Systematic 

Dissection of the Unfolded Protein Response. Cell 167, 1867–1882.e21 (2016). 

8. Jaitin, D. A. et al. Dissecting Immune Circuits by Linking CRISPR-Pooled Screens with 

Single-Cell RNA-Seq. Cell 167, 1883–1896.e15 (2016). 

9. Liu, Z., Lavis, L. D. & Betzig, E. Imaging live-cell dynamics and structure at the single-molecule 

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted August 27, 2019. . https://doi.org/10.1101/747758doi: bioRxiv preprint

level. Mol. Cell 58, 644–659 (2015). 

10. Li, N. et al. Single-molecule imaging and tracking of molecular dynamics in living cells. Natl Sci 

Rev 4, 739–760 (2017). 

11. Balzarotti, F. et al. Nanometer resolution imaging and tracking of fluorescent molecules with 

minimal photon fluxes. Science aak9913 (2016). 

12. Liu, T.-L. et al. Observing the cell in its native state: Imaging subcellular dynamics in 

multicellular organisms. Science 360, (2018). 

13. Baltekin, Ö., Boucharin, A., Tano, E., Andersson, D. I. & Elf, J. Antibiotic susceptibility testing 

in less than 30 min using direct single-cell imaging. Proceedings of the National Academy of 

Sciences 201708558 (2017). 

14. Hammar, P. et al. Direct measurement of transcription factor dissociation excludes a simple 

operator occupancy model for gene regulation. Nat. Genet. 46, 405–408 (2014). 

15. Wang, P. et al. Robust growth of Escherichia coli. Curr. Biol. 20, 1099–1103 (2010). 

16. Lawson, M. J. et al. In situ genotyping of a pooled strain library after characterizing complex 

phenotypes. Mol. Syst. Biol. 13, 947 (2017). 

17. Wallden, M., Fange, D., Lundius, E. G., Baltekin, Ö. & Elf, J. The Synchronization of 

Replication and Division Cycles in Individual E. coli Cells. Cell 166, 729–739 (2016). 

18. Si, F. et al. Mechanistic Origin of Cell-Size Control and Homeostasis in Bacteria. Curr. Biol. 0, 

(2019). 

19. Ghatak, S., King, Z. A., Sastry, A. & Palsson, B. O. The y-ome defines the 35% of Escherichia 

coli genes that lack experimental evidence of function. Nucleic Acids Res. 47, 2446–2454 (2019). 

20. Ranefall, P., Sadanandan, S. K. & Wählby, C. Fast adaptive local thresholding based on ellipse 

fit. in 2016 IEEE 13th International Symposium on Biomedical Imaging (ISBI) 205–208 (2016). 

21. Magnusson, K. E. G., Jalden, J., Gilbert, P. M. & Blau, H. M. Global linking of cell tracks using 

the Viterbi algorithm. IEEE Trans. Med. Imaging 34, 911–929 (2015). 

22. Olivo-Marin, J.-C. Extraction of spots in biological images using multiscale products. Pattern 

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted August 27, 2019. . https://doi.org/10.1101/747758doi: bioRxiv preprint

Recognit. 35, 1989–1996 (2002). 

23. Jaqaman, K. et al. Robust single-particle tracking in live-cell time-lapse sequences. Nat. Methods 

5, 695–702 (2008). 

24. Keyamura, K. et al. The interaction of DiaA and DnaA regulates the replication cycle in E. coli 

by directly promoting ATP DnaA-specific initiation complexes. Genes Dev. 21, 2083–2099 

(2007). 

25. Katayama, T., Kasho, K. & Kawakami, H. The DnaA Cycle in Escherichia coli: Activation, 

Function and Inactivation of the Initiator Protein. Front. Microbiol. 8, 2496 (2017). 

26. Heller, D. M., Tavag, M. & Hochschild, A. CbtA toxin of Escherichia coli inhibits cell division 

and cell elongation via direct and independent interactions with FtsZ and MreB. PLoS Genet. 13, 

e1007007 (2017). 

27. Saxena, R., Fingland, N., Patil, D., Sharma, A. K. & Crooke, E. Crosstalk between DnaA protein, 

the initiator of Escherichia coli chromosomal replication, and acidic phospholipids present in 

bacterial membranes. Int. J. Mol. Sci. 14, 8517–8537 (2013). 

28. Camara, J. E. et al. Hda inactivation of DnaA is the predominant mechanism preventing 

hyperinitiation of Escherichia coli DNA replication. EMBO Rep. 6, 736–741 (2005). 

29. Fujimitsu, K., Senriuchi, T. & Katayama, T. Specific genomic sequences of E. coli promote 

replicational initiation by directly reactivating ADP-DnaA. Genes Dev. 23, 1221–1233 (2009). 

30. Kasho, K. & Katayama, T. DnaA binding locus datA promotes DnaA-ATP hydrolysis to enable 

cell cycle-coordinated replication initiation. Proc. Natl. Acad. Sci. U. S. A. 110, 936–941 (2013). 

   Acknowledgments This work was supported by the Knut and Alice Wallenberg foundation (KAW) and the Swedish research council (VR). We are grateful to Dr. Irmeli Barkefors for help with the manuscript and figures and to Praneeth Karempudi for making microfluidic molds.   Author contributions JE conceived the DuMPLING method and SeqA application, DC developed cloning methods, designed and made strain library, JE and ML managed the project, DF and ML developed phenotyping methods, JL and ML developed genotyping methods, DF built the 

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted August 27, 2019. . https://doi.org/10.1101/747758doi: bioRxiv preprint

microscope, JL and ML made microscopy experiments, SZ, DF developed image analysis pipeline, SZ, DF analysed DuMPLING data, DJ made repression measurements and NGS growth rate experiments, JE, ML, DC and DF wrote the paper with input from all authors.   Data and code All raw data and analysis codes will be publically deposited at the time of publication.          

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted August 27, 2019. . https://doi.org/10.1101/747758doi: bioRxiv preprint